Learning movements from a virtual instructor

Abstract

We examined the effects of perspective (first person versus third person) and immersion (immersive versus nonimmersive) on motor learning in order to assess the format of action representations. Participants viewed the instructor from either a first or a third person perspective. During immersive conditions, they wore a 6 DoF-tracked head-mounted display, as in Figure 1(a). For nonimmersive conditions, they viewed a computer monitor, as in Figure 1(b). We also evaluated whether these effects were modulated by experience. Experienced dancers and novices practiced dances by imitating a virtual instructor and then subsequently had to perform the dances from memory without an instructor present, following a delay. Accuracy for both practice and test trials was video coded. In line with theoretical models of motor learning, mean accuracy increased with successive trials in accordance with the power law of practice. First person perspective formats led to better accuracy, but immersive formats did not, as shown in Figure 2. Experienced dancers were more accurate than novices, but format did not interact with experience. These results suggest that during learning, individuals across experience levels represent complex actions in first person perspective, and that virtual instruction does not require immersion to be effective.